Super-resolution utilizing total variation regularization and a shock filter
Masaru Sakurai, Y. Sakuta, Masashi Watanabe, Tomio Goto, Satoshi Hirano · 2012
The total variation (TV) regularization method is a very effective technique for super-resolution image reconstruction, however, traditional TV super-resolution approaches require a lot of computational time. In this paper, we propose a novel system in which a shock filter is introduced into the TV up-sampling system. The result obtained is very effective both in terms of picture quality and computational time. With respect to picture quality, the edge sharpness has been much improved compared with the conventional TV up-sampling method. While with regard to the computational time, it has been significantly shortened and the possibility of the application of our system for motion images comes into view. We have examined the feasibility of the application of our system to motion pictures by implementing the system on a CELL processor, and have proved that SDTV-HDTV conversion is possible in real time.